Browser
Use the headless browser that comes with every computer.
Every computer comes with Lightpanda, a headless browser made for agents and automation. It's small and quick: it uses about 11 MB of memory when idle, where Chrome needs hundreds.
Fetch a page
lightpanda fetch --dump html https://example.comThis loads the page, runs its JavaScript, and prints the resulting HTML.
Drive it with Puppeteer or Playwright
Start it as a server, then connect the way you'd connect to Chrome:
lightpanda serve --host 127.0.0.1 --port 9222 &// Puppeteer
const browser = await puppeteer.connect({ browserWSEndpoint: "ws://127.0.0.1:9222" });
// Playwright
const browser = await chromium.connectOverCDP("ws://127.0.0.1:9222");Node.js isn't installed by default: sudo apt-get install -y nodejs npm.
What it can't do
- No screenshots. It doesn't draw pages, so agents that work by looking at the screen can't use it. Agents that read the page's HTML or text can.
- Some sites break. It doesn't support every web feature yet, so a few sites load partly or not at all. If one does, try again or use Chrome.
Need full Chrome?
sudo apt-get install -y chromium
chromium --headless --screenshot=page.png https://example.comChrome handles every site and can take screenshots, but it needs much more memory. On a 4 GB computer, keep it to a few tabs.
Privacy and licence
Lightpanda normally sends usage statistics to its makers. We turn that off on every computer (LIGHTPANDA_DISABLE_TELEMETRY=true). It's used unmodified under the AGPL-3.0 licence; the licence and a link to its source are in /usr/share/doc/lightpanda.